关于MLMS算法步长的一点注记

NOTE TO THE STEPSIAE OF MLMS ALGORITHM

用数学分析和估值理论的方法研究修正ＬＭＳ算法的收敛性能与步长之间的关系，导出算法收敛的必要条件和性能失调与步长之间的约束条件；将修正ＬＭＳ算法与ＬＭＳ算法比较，得到修正算法实质上是有界的结论，无论修正算法中的参数μ取值多大，其实际计算步长都小于ＬＭＳ算法步长的上限。实验结果与理论分析相符合。

The necessary condition of the conyergence of the modified least mean square algorithm(MLMS)and the restrictive condition between the performance misadjustment and the stepsize are derived by studying the relationship of the mathematics analysis and the estimation theory. The conclusion that the stepsize of MLMS agorithm is finite in na-ture comes from the comparison of the MLMS algorithm to the LMS algorithm. However big the parameter μ is,the upper bound of the stepsize in the MLMS algorithm is always less than that in the LMS algorithm. The experimental result corresponds to the above theory.